On Friday shares of Intel (INTC), a long-standing favorite of ours, dropped by over 5 percent. The catalyst for the drop was Intel's announcement of a big hike in its capital expenditures, to $13 billion, from an expected $11 billion.
Our contrarian nature makes us suspicious of any company pummeled on news of higher capital expenditures. After all, the major expenditures only seemed to affirm Intel's need to stay ahead of its competition in semiconductor manufacturing. But many events over the weekend and past several weeks have begun to persuade me that Intel, and indeed most of the companies in the information technology space, are at the end of - or close to the end of - a long period of dramatically reducing costs. More important, these same events reawakened my long held view that computers are tools and only tools… and are insufficient or probably even unnecessary for further human progress. If I am correct, we may be at the end of an era.
The implications of an end to ever-faster computers are far more profound than our waning regard for Intel's clearly continuing prowess. The end (or near end) of IT improvements could reverberate through our economy much like scarce energy and the growing scarcity of other critical resources, potentially thereby lowering productivity and raising costs for many products.
Inflation - and in particular, core inflation - has largely been kept in check by technical gains and increasing IT cost-efficiencies that offset the impact of rising prices in other arenas. But if no more IT improvements can be easily squeezed from semiconductors, prices would be likely to rise, and to rise sharply.
Should the semiconductor industry hit this wall, we could accept IT limitations (along with resource scarcity) as challenges and realize that to improve our condition we must rely to a much greater extent on our own human intuition. Clearly, I hope humans will increasingly deploy our intuition, even if cost efficiencies continue to improve. And there is no shortage of areas in which human intuition could indeed lead to the kinds of breakthroughs we have long dreamt about. Quantum computing (which is not beyond the capabilities of today's IT infrastructure), and better understanding of exotic materials like graphene come immediately to mind out of a gigantic list.
The events that led to my change in thinking derive first from my devotion to the game of chess. I admittedly stand far from being a serious competitive chess player, but I love to analyze games. And indeed, every week I meet with a friend and top chess coach to do exactly that - review the latest great chess games and sometimes much earlier ones (i.e., from the pre-computer era). In recent years, my interest in chess was rekindled by the widely accepted belief that computers are clearly better players than humans at the game of Kings. I don't buy that - or at least I don't buy it completely. More than 15 years have passed since then-world champion Gary Kasparov lost a chess match to a computer dubbed "Deeper Blue." We pointed out in our book Defying the Market that Kasparov's loss had much more to do with human nerves and frailties than a lack of intellectual superiority.
In the ensuing years computer speeds have grown immensely. Today a high-powered notebook computer can easily match the power of Deeper Blue. And while computers can beat many humans, several aspects of the game remain well beyond the capabilities of computers -clearly demonstrating the superiority of over computers of human intuition. In one game I watched, a young super master had arrived at a poor position. The best computer program analyzed the game and suggested several moves - all of which led to a weak, albeit not necessarily losing position. But then the young wunderkind made a move not even considered by the computer. Once he made this move, the computer immediately changed its evaluation from poor to losing - and maintained that evaluation for at least two or three moves. However, it became clear that the human move was the only one that would have achieved a draw, and that was indeed the final result. This was not an isolated example. In countless instances, humans "see" something well beyond the reach of the best computers. The examples from chess make it clear that, despite vast improvement in computer speed, humans still do the computer programming - and ironically, the likely future limitations we will face in computer hardware need not also ring a death knell for human progress. They almost certainly will not. Indeed, further improvements in chess playing computers must come from humans developing the chess programs. Faster computers (i.e., speed alone) simply cannot trump human creativity. In the same way, to believe that computers are the be-all and end-all answer to the world's woes would be a terrible mistake.
The second event that reordered my thinking was a Science Magazine obituary of world-renowned biologist Carl Woese. He did his great work in the 1970's, long before the advent of computers made DNA sequencing a relatively simple task. Working by hand, Woese was able to distinguish a third branch on the tree of life. His feat involved both great intuition and long, long hours. I have felt for some time that we just don't give ourselves enough credit when comparing our brains to computers. Perhaps reaching fundamental limitations in computer speeds will force us to face the incredible potency of the human mind, especially a human mind has a powerful computer as a tool - and only a tool - to extend its reach.
Limitations in chip size and computer speeds don't arrive due to our reaching a theoretical end to Moore's law. This states that transistor size shrinks by about half every 18-24 months (so, therefore, computer speed increases by a similar factor in the same period). Of course, there is a theoretical point after which transistors are limited by fundamental physical factors. But we think more immediate limitations stem from more prosaic monetary factors. To leverage new semi-conductor designs, larger wafers are needed to make costly changes on a larger scale. In other words, the larger the wafer on which we fit the circuits, the greater the number of chips we can process at once. If wafer size does not change, costs will rise exponentially, along with complexity. Fabrication plants that worked on smaller wafers cannot be retrofitted to process larger wafers. The catch is that the cost of such plants rises dramatically with size. The current wafer size is 300 millimeters and the cost of a fabrication plant to process this size runs from $4 billion to $10 billion. Now, plans are afoot to move to wafer sizes to 450 millimeters to accommodate much smaller circuits.
Alas, these 450 mm fabrication plants will likely cost more than $10 billion (and very possibly much more), something closer to the price of a nuclear reactor. Currently, only three companies can potentially afford these plants - Intel, Samsung, and Taiwan Semiconductor. And given the complexity associated with these new circuits, it is not clear that the massive capital needed to build the plants will yield a positive return. Moreover, even if Intel et al. succeed it looks pretty clear that we have reached the end of the line. Beyond a wafer size of 450 mm, costs are likely to far exceed the financial capacity of any company. Not surprisingly, the industry trend is to make use of older plants to produce simpler designs. More and more, the goal of the semiconductor industry will be "not faster, but better." (And this is where human intuition is at a huge premium.)
There may still be a place in your portfolio for Intel, but we no longer consider it a must-own - even for technology investors. Rather, we feel there will be a shift toward IT consulting firms. Such companies advise other firms how to best utilize their hardware and which software to use. Two favorites in this category are IBM (IBM) and Accenture (ACN). Both (and IBM in particular) are multifaceted service companies that also manufacture technology products and outsource client company functions. Their metrics appear compelling. Largely unencumbered by heavy capital requirements, they are full-fledged cash machines, with high free cash flow yields, which translate into steady share buy backs and rising dividends. Clearly, there are a massive number of such service and consulting companies, which will continue to make more and more out of relatively fixed processing speeds.
The bottom line is that success is in our own hands now, and not guaranteed in any way by ever faster machines. Let's not drop the ball.